Rainfall induced soil slips are among the most dan-gerous natural hazards acting on hillslopes, leadingto structural damage and casualties. These shallowlandslides are triggered by periods of intense pre-cipitation, very often falling on already wet soils. Inmost cases the failures can be considered transla-tional mass movements that occur suddenly, practi-cally giving no time to warn the communities at risk.The best available prevention tool for large areasmust therefore rely on the ability to predict the timingand location of these natural hazards well in ad-vance.

Switzerland is particularly prone to rainfall trig-gered soil slips, given the steep nature of many of itsalpine and pre-alpine slopes. A series of natural dis-asters involving soil slips affected several areas inthe country in the last decade, among these are theones occurred in the region around Napf (Emmentalarea) in July 2002. The event was accurately docu-mented (Rickli & Bucher 2003) by the Swiss FederalInstitute for Forest, Snow and Landscape Research(WSL). The site was chosen here with the aim of as-sessing the predictive capability of a time dependentslope stability model developed by Baum et al.(2002).

The model, TRIGRS (Transient Rainfall Infiltrationand Grid-based Slope-stability), is a coupled hydro-mechanical slope stability assessment tool, workingon a regional scale. TRIGRS is raster based, anduses a time-dependent approach to assess the sta-bility of a basin during a rainfall event. Infiltration ismodelled through a simplified analytical solution ofRichards’ equation (Iverson 2000), which requires ashallow, quasi-saturated soil cover at the beginningof a simulation. The built-up pore pressure is thenused as an input to a slope stability model, based onthe infinite-slope approach. The main output ofTRIGRS is a factor of safety, indicating whether theslope can be considered stable or not. The assump-

tion that the soil must be quasi-saturated at the be-ginning of a simulation can be viewed as a limitationof the model for soils that are more than a couple ofmeters deep, but is not an issue in the case exam-ined. The soil cover in the test area does not exceed3 meters even in the flatter regions.

The main problem in applying TRIGRS is the lackof detailed input data for topography and soil overlarge areas. A new high-resolution DTM (Digital Ter-rain Model) with an average density of 0.5 points/m2

was made available by the cantonal authorities, andwas first translated into a TIN (Triangulated IrregularNetwork), using the standard triangulation procedureprovided by the software ArcGIS. It was then con-verted into a grid with a resolution of 3 meter usingArcGIS. TRIGRS proved to be very sensitive to theslope angle, making the availability of a precise DTMa primary need, but also presenting the problem ofchoosing a reliable routine to determine the slopeangle for each cell of the raster. Given that the re-sulting 3 meter-grid seemed to be affected by thepresence of artefacts (namely terraces), ArcGis to-pographic slope, a notoriously smoothing routine,was chosen to compute the slope angle. It must benoted that even with a high resolution DTM none ofthe examined slope computation procedures wasable to match the measured slope angles in the field.This problem is also addressed by Rickli & Bucher(2003) for a lower resolution DTM.

A series of hypotheses were made to model thevariation of some selected TRIGRS input parame-ters. These conceptualizations were necessary dueto the lack of information about the variability ofthese parameters in the study area. Cohesion wasmade dependent on altitude, relying on the assump-tion that soils at lower elevations are less disturbedat the base (where the lowest factor of safety isfound) than the shallower ones at higher elevations.Soil depth was defined as a decreasing exponential

Predicting rainfall triggered soil slips:

a case study in the Emmental region (Switzerland)

Bisanti, B., Molnar, P. & Burlando, P.

Institute of Hydromechanics and Water Resources Management, Swiss Federal Institute of Technology (ETH), Zurich,Switzerland

function of the slope angle, ranging from about 1meter on the steeper slopes to about 3 meters in thevalley. The initial water table height was describedas a function of the slope angle. Average values,computed from the few available data, were used forthe friction angle, the saturated hydraulic conductiv-ity and related hydraulic diffusivity. Hourly rainfalldata for the meteorological station Napf was ob-tained from MeteoSwiss and distributed uniformlyover the basin. used.

The preliminary results show the ability of themodel to capture 41% of the occurred slides by de-stabilizing 12.6% of the entire basin (Figure 1). Thetendency to overestimate instabilities on steep areasis still present, and seems rather unavoidable withthe available data. An improvement in this sensecould be expected if additional information on lan-duse for the area under investigation can be ob-tained, for instance through the use of aerial photo-graphs. The model does not in fact distinguishbetween different landcover areas, such as forest,meadows, etc.

Nevertheless TRIGRS holds good promises as aprediction tool for rainfall induced soil slips, butneeds to be further tested in other areas, with spe-cial attention devoted to the conceptual assumptionsmentioned above.

Figure 1. The picture shows the study area at the end of a 3-hour simulation. The white dots represent actual landslide loca-tions, while the dark grey areas are the simulated unstablezones as output by TRIGRS.

REFERENCES

Iverson, R.M. (2000): Landslide triggering by rain infiltration.Water Resources Research, 36 (7): 1897-1910.

Baum, R.L., Savage, W.Z. & Godt, J.W. (2002): TRIGRS-AFortran Program for Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Analysis. U.S. GeologicalSurvey Open-Fi le Repor t 02-0424, 27p.http://pubs.usgs.gov\of\2002\ofr-02-424

Rickli, C. & Bucher, H. (2003): Oberflächennahe Rutschungen,ausgelöst durch die Unwetter vom 15.-16.7.2002 imNapfgebiet und vom 31.8.-1.9.2002 im Gebiet Appenzell -Projektbericht zuhanden des Bundesamtes für Wasser undGeologie BWG.

The Monte Rosa east face, Italian Alps, is one of thehighest flanks in the Alps (2200 – 4500 masl). Steephanging glaciers and permafrost cover large parts ofit. Since the end of the Little Ice Age (about 1850),the hanging glaciers and firn fields have retreatedcontinuously. During the recent decades, the icecover of the Monte Rosa east face experienced anaccelerated and drastic loss in extent. Some glaciershave totally disappeared leaving large parts of theunderlying rock unprotected against mechanical andthermal erosion. Enhanced rock fall and debris flowactivity was observed (Haeberli et al., 2002; Kääb etal., 2004; Fischer, 2004). Perennially frozen rockwalls are highly complex systems that may reactvery sensitively on changes. Those changes in gla-cier extent, permafrost conditions, thermal and hy-drological regimes, as related to present atmos-pheric warming, significantly affect the stabilityconditions of the Monte Rosa east face.

The exceptional rock fall activity during the hotsummer 2003 has pointed to the relation of rock falland climate change via permafrost thaw. These nu-merous smaller events stemmed mainly from en-larged active layer thickness during this extraordi-nary summer. Also the most recent event in theMonte Rosa east face, a 0.5 to 1 Mio. m3 largeice/rock avalanche in August 2005, underlines theongoing development of instabilities in ice and rock.

The scope of this study is to analyze the linkagebetween the glacier shrinkage and permafrost deg-radation, on the one hand, and the observed in-creasing slope instabilities in the Monte Rosa eastface, on the other hand (Fischer, 2004). A number ofamateur photos, air-photos and maps was compiledin order to reconstruct the development of the icecover of the Monte Rosa east face. The geology ofthe Monte Rosa east face and the detailed extents ofthe hanging glaciers were mapped during fieldwork

in summer 2003. The starting zones of rock fall, iceavalanches and debris flows were observed and lo-calized as well. The permafrost distribution in therock wall was computed with different models.Roughly one half to two thirds of the east face areestimated to be under permafrost conditions.

For the compilation and processing of these data,a Geographic Information System (GIS) was used.GIS technologies facilitate the integration of remotesensing and field data as well as modelled data foranalysing and modelling instable and hazardouszones in a steep slope.

The investigated parameters are shown as sepa-rate layer in the GIS. The hazardous areas can bedetected and classified by an overlay and intersec-tion of the different layers.

It turned out that:- Most of the active starting zones of rock fall and

debris flow are located in parts of the rock wall,where surface ice disappeared recently.

- Most of the active starting zones are located inpermafrost zones, mostly close to the estimatedlower boundary of the permafrost occurrence.

- Many active starting zones are situated at theboundaries between two different lithologies.

In the view of ongoing or even enhanced atmos-pheric warming it is therefore very likely that the in-stabilities in the Monte Rosa east face will continueto represent a critical hazard source. Thereforesome first-order modelling of rock fall events and iceavalanches has been conducted showing that par-ticularly large events could endanger some parts ofthe upper part of the Valle Anzasca, especially in thecurrent situation of an elevated Ghiacciaio delBelvedere and an occasionally filled supraglaciallake on it.

Geology, glacier changes, permafrost and related slope instabilitiesin a high-mountain rock face:

Monte Rosa east face, Italian Alps

Fischer, L., Kääb, A., Huggel, C. & Nötzli, J.

Glaciology and Geomorphodynamics Group, Department of Geography, University of Zurich, Switzerland

This study shows the high suitability of the GIStechnology for the processing and presentation ofmulti-temporal and multi-spatial data and relatedprocesses. It reveals as an important tool for theevaluation and prediction of natural hazards in gen-eral and the detection of possibly instable and haz-ardous zones in high alpine rock walls in detail.

REFERENCES

Fischer, L. (2004): Monte Rosa Ostwand – Geologie, Ver-gletscherung, Permafrost und Sturzereignisse in einerhochalpinen Steilwand. M.Sc.-Thesis, Department of Geog-raphy, University of Zurich.

Haeberli, W., Kääb, A., Paul, F., Chiarle, M., Mortara, G.,Mazza, A., Deline, P. and Richardson, S. (2002): A surge-type movement at Ghiacciaio del Belvedere and a devel-oping slope instability in the east face of Monte Rosa,Macugnaga, Italian Alps. Norwegian Journal of Geography56, 104-111.

Kääb, A., Huggel, C., Barbero, S., Chiarle, M., Cordola, M.,Epifani, F., Haeberli, W., Mortara, G., Semino, P., Tam-burini, A. and Viazzo, G. (2004): Glacier hazards atBelvedere glacier and the Monte Rosa east face, ItalianAlps: Processes and mitigation. International Symposium,Interpraevent 2004 – Riva/Trient.

Permafrost degradation has been hypothesized anddemonstrated to influence rock-wall stability. Boththaw and warming of permafrost (entering the rangeof -1.5 to 0 °C) as well as the build-up of hydrostaticpressure following thaw are possible mechanismsthat link warming to the reduction in strength of ice-bonded rock joints (Haeberli et al. 1997, Davies et al.2001, Gruber et al. 2004a).

Quantitative information on the spatial distributionof this additional, warming-related stability factor isdesirable to support the assessment of natural haz-ards in mountain areas. This contribution proposesvariables that describe this effect and explores a wayto account for the large uncertainty inherent in theirmodelling.

The most important variable to delineate zones ofpossible rock fall induced by permafrost degradationis the occurrence of permafrost underneath a sur-face. The degradation of a permafrost body will takeplace along its boundary. This can be at the perma-frost table, the permafrost base and also result fromlateral heat fluxes in complex topography. The depthof the degrading boundary of the permafrost bodycorresponds to the magnitude of a rock fall inducedby thaw. The additional heat flow at the boundarycorresponds to the frequency or likelihood of anevent taking place as a consequence of warming asit is proportional to the volume of material that canbe warmed or the volume of ice melted (Fig. 1).

For any given warming scenario, the resulting fre-quencies and magnitudes change over time. Espe-cially for regional-scale modeling sub-surfacethermo-physical properties and water/ice contentsare unknown and can thus vary in a wide range andinfluence the subsurface temperature field accord-ingly. Additionally, the surface temperature boundarycondition simulated by energy-balance models has ahigh uncertainty in complex topography. Uncertain-

ties of driving temperature scenarios further add tothis for future projections. The uncertainties of theseeffects can be propagated using Monte-Carlo tech-niques.

Figure 1. Schematic of proxies for the frequency and magni-tude of instabilities induced by permafrost degradation.

In this investigation we explore the calculation andinterpretation of the proposed proxies together withuncertainty propagation techniques. Rock tempera-tures for south- and north-facing locations and di-verse elevations were simulated with the model TE-BAL (Gruber et al. 2004b) based on hourlymeteorological data. The uncertainty of four key pa-rameters was propagated. Volumetric heat capacity,thermal conductivity and water content of the rock aswell as an assumed error of the simulated surfacetemperature were sampled along assumed probabil-ity density functions during Monte-Carlo simulationswith 150-500 realizations per point. These parame-ters were sampled from normal distributions trun-cated at 3σ. No stratification and no change in totalwater content was assumed for the sub-surface. Astandard random sampling scheme was employedwithout dedicated improvements such as latin hyper-cube sampling. Temperature profiles were initializedwith the mean surface temperature of 1990-1993and then spun with 1993 data to achieve a realistictemperature profile. During the simulation, the years1993-2002 were taken as the baseline run that the

Deriving proxy variables for frequency and magnitude of rock fallinduced by permafrost thaw using Monte-Carlo simulation of

surface and sub-surface heat transfer

1,2 Gruber, S., 2 Noetzli, J. & 3 Kohl, T.

1 Laboratoire EDYTEM, Université de Savoie, France2 Glaciology and Geomorphodynamics Group, Department of Geography, University of Zurich, Switzerland

3 Geowatt AG, Zurich, Switzerland

simulated temperatures from the extreme year 2003were compared to.

From the modelled transient temperature fieldsthree quantities were extracted: 1) probability ofpermafrost occurrence based on the percentage ofrealizations with permafrost in the baseline run; 2)active layer depth based on the depth of 0 °C inmaximum temperature of baseline; and 3) excessheat content of the permafrost body defined as theadditional heat content (including latent heat) at aspecific date compared with the maximum during thebaseline run. From these results, the active layerdepth not exceeded with a certain probability (Fig. 2)can be derived from its cumulative frequency (eitherfor the baseline or 2003). The excess heat content ofthe permafrost body (proxy for frequency of likeli-hood of destabilization) was calculated for severaldates during 2003 and expressed in mm waterequivalent that could have been melted in order toarrive at numbers that can readily be put into con-text.

Figure 2. Cumulative probability distribution of the active layerdepth (magnitude proxy) for several elevations, north-facingduring the baseline run 1993-2002.

The findings of earlier modelling (Gruber et al.2004a) were reproduced: The timing of observedrockfall and modelled destabilization conditions doesnot match. While much rock fall in 2003 took place atlow altitudes and early in the year simulations resultssuggest destabilization in late summer or autumn forthose locations. This points towards important chal-lenges for future research about the processes thatconnect warming and permafrost degradation withthe destabilization of rock slopes. Ice segregationdue to the migration of unfrozen water along thermalgradients (Murton et al. 2001) is a process that maybe underestimated in this context. Since it is driven

by temperature gradients it may act earlier in theyear that maximum temperatures or heat contents.

In contrast to the experiment presented here,permafrost bodies in mountain environments arecomplex and heterogeneous 3-dimensional objects.For the future, not only active-layer thickening butalso long-term changes in the geothermal field needto be taken into account in the context of tempera-ture-related stability changes (Fig. 3).

Figure 3. Schematic of possible future destabilization proxies.

The propagation of uncertainty in such modelswill be a key future step because it gives importanceand visibility to the unknown instead of presentingaccuracy in selected example cases. Findings basedon simulations results can thus be interpreted andtrusted better. However, such a development shouldnot lead to the illusion of “certainty of uncertainty”.Especially the determination of the probability distri-bution of input parameters will continue to be chal-lenging and require careful interpretation of results.

REFERENCES

Davies, M. C. R., Hamza, O. & Harris, C. (2001): The effect ofrise in mean annual temperature on the stability of rockslopes containing ice-filled discontinuities. Permafrost andPeriglacial Processes 12: 137-144.

Gruber, S., Hoelzle, M. & Haeberli, W. (2004): Permafrost thawand destabilization of alpine rock walls in the hot summer of2003. Geophysical Research Letters 31: L13504.

Gruber, S., Hoelzle, M. & Haeberli, W. (2004): Rock-wall tem-peratures in the Alps: modelling their topographic distribu-tion and regional differences. Permafrost and PeriglacialProcesses 15(3): 299-307.

Haeberli, W., Wegmann, M. & Vonder Mühll, D. (1997): Slopestability problems related to glacier shrinkage and perma-frost degradation in the Alps. Eclogae geol. Helv. 90: 407-414.

Murton, JB., Coutard, J-P., Ozouf, J-C., Lautridou, J-P., Robin-son, D.A. & Williams, R.B.G. (2001): Physical modelling ofbedrock brecciation by ice segregation in permafrost. Per-mafrost and Periglacial Processes 12: 255–266.

Due to the difficulty of describing the complex spatialand temporal patterns inherent to volcanism, the useof solely deterministic models is not sufficient forlong term estimation of volcanic hazards. In order toaccount of the intrinsic uncertainty of volcanism thatoccurs in space, time and with respect to event typesand their intensity, the use of probabilistic modelsbecomes quite natural for long term hazard assess-ment.

The motivation of this on-going work is driven bythe necessity of investigating various probabilisticapproaches for the estimation of volcanic hazards inrelation to repository siting in Japan. In particular, theintegration of additional sources of uncertaintieslinked to the distribution of volcanic events in spaceas well as to their occurrence need to be accountedfor by the modelling. The concepts of the geostatisti-cal models of the proposed approach will be pre-sented and then illustrated by a case study usingdata from the Tohoku volcanic arc (Fig.1).

Figure 1. Data from the Tohoku volcanic arc

Geostatistical models for the long term estimationof volcanic hazards

Jaquet, O.

Colenco Power Engineering, Baden, Switzerland

Deposits of simultaneously occurring sublacustrinemass movements in lakes of Central Switzerland areused as paleoseismological archives to identify theintensity, epicentral location and recurrence rates ofstrong prehistoric earthquakes. Three major paleo-seismic events were identified in the past 15,000years that were strong enough to trigger multiplesublacustrine slope failures in Lake Lucerne as wellas in Lake Zurich (Macroseismic local intensity I >=VII). This indicates that strong earthquakes can af-fect a big area in Central Switzerland, where suchevents are generally not expected.

Previous studies have shown that sublacustrineslope failures in lakes of Central Switzerland can beattributed to historic earthquakes. A study comparingthe lacustrine fingerprints of four moderate to stronghistoric earthquakes with moment magnitudes of Mw

= 5.7 to Mw = 6.9 and epicentral intensities of I0=VIIto I0=IX reveals that subaqueous mass movementsonly occurred in the studied lakes (Lake Sarnen,Lake Lungern, Lake Lucerne, Lake Baldegg, andSeelisberg Seeli) if they are situated within an areathat underwent groundshaking not smaller thanintensity VII (Monecke et al., 2004). New results froma limit equilibrium back analyses for sublacustrineslope stability in Lake Lucerne also confirm that themodelled lacustrine slope is stable under staticloading conditions. Slope failure only occurs if theseismic ground accelerations exceed 0.75 ms-2.Quasi-3D high-resolution geophysical imaging (3.5kHz pinger source) of the subsurface, ground-proofed with a series of sediment piston coresreveals the spatial and temporal distribution of LateGlacial to Holocene mass movement deposits inLake Lucerne and Lake Zurich. The detailed basin-wide 3D stratigraphic correlations in combinationwith accurate radiocarbon and tephrochronologicaldating allows identifying synchronously occurring

multiple mass movement deposits that can be usedas paleoseismic indicators. For an example, thehistorically well described 1601 A.D. Unterwaldenearthquake (I0 = VII–VIII / Mw ~ 6.2) triggerednumerous synchronous mass movements andmegaturbidites within different basins of LakeLucerne, producing a characteristic pattern that canbe used to assign a seismic trigger mechanism toprehistoric mass movement events (Schnellmann etal. 2002). Furthermore, this historic event showedthat seismic hazard for lakeshore communities isamplified by slide-induced tsunami and seichewaves that reached wave heights up to 4 m. In LakeLucerne, beside the historic 1601 A.D. Unterwaldenearthquake, five prehistoric events occurred at 2300+/- 120, 9870 +/- 45, 11’730 +/- 255, 13’710 +/- 130,and 14590 +/- 830 calendar yr B.P (Schnellmann etal, submitted).

Results of a new study indicate that such multipleslope failures are not limited to inner-alpine lakes butthat they do also occur in Lake Zurich, where threemultiple subaqueous mass movements wereidentified and dated to 2150 +/-100, 11’770 +/- 220and 14’000 +/- 450 cal. yr B.P. Within theradiocarbon dating uncertainties, all three LakeZurich events occurred simultaneously with multiplemass movement events recorded in the subsurfaceof Lake Lucerne. Our data therefore reveal that threestrong paleoseismic events occurred in the last15’000 years in Central Switzerland, which werestrong enough to trigger multiple subaqueous massmovements in the two different lakes that are ~40km apart.

The 1601 A.D. Unterwalden earthquake was notregistered in Lake Zurich. It was attributed to an epi-central area in the Helvetic domain south of thenorthern Alpine front (Sarnen, Unterwalden) that is

Major prehistoric earthquakes along the northern alpine frontrevealed by simultaneous slope failures

in Lake Zurich and Lake Lucerne

*Strasser M., **Schnellmann M., ***Monecke K., *Bussmann F., *Anselmetti, F.S. & ****Giardini, D.

* Geological Institute, ETH Zurich, Switzerland, strasser@erdw.ethz.ch

** Genossenschaft für die Lagerung radioaktiver Abfälle (NAGRA), Wettingen, Switzerland

*** Sedimentology Section, Department of Surface Waters (SURF), EAWAG Dübendorf, Switzerland

**** Geophysical Institute, ETH Zürich, Switzerland

known to have been seismically active during histori-cal times (Schwarz-Zanetti et al. 2003). For the pre-historic events recorded with macroseismic intensi-ties of > VII in both, Lake Zurich and Lake Lucerne,a similar earthquake source area evidently wouldhave been to far away to trigger slope failures inLake Zurich situated ~50km NE of the 1601 A.D.epicenter. As a possible source area for these majorevents we postulate an earthquake trigger sourcesomewhere along the northern alpine front, wheremajor thrust faults still might be seismically active.This scenario would imply major earthquakes alongthe northern alpine front with recurrence rates lessthan 10’000 years.

REFERENCES

Monecke, K., Anselmetti, F.S., Becker, A., Sturm, M. andGiardini, D. (2004): The record of historic earthquakes inlake sediments of Central Switzerland. Tectonophysics,394(1-2): 21-40.

Schnellmann, M., Anselmetti Flavio, S., Giardini, D., McKenzieJudith, A. and Ward Steven, N. (2002): Prehistoric earth-quake history revealed by lacustrine slump deposits. Geol-ogy, 30(12): 1131-1134.

Schnellmann, M., Anselmetti Flavio, S., McKenzie Judith, A.and Giardini, D., submitted to Eclogae Geologicae Helvetia.Spatial and temporal distribution of late quaternary massmovements in and around lake Lucerne.

Schwarz-Zanetti, G. et al. (2003): The earthquake in Unter-walden on September 18, 1601: A historico-critical macro-seismic evaluation. Eclogae Geologicae Helvetiae, 96(3):441-450.

Radon 222 (Rn222) is a natural radioactive gas re-sulting from the decay of U238. It can reach high con-centration values inside closed environments thatare exposed to constant emanations and can exceedmaximum acceptable exposure levels. Recent stud-ies (Wichmann 2005) on trends among lung cancercases has determined significant health threads ofdeveloping cancer from the 140 Bq/m3 level, due tolong periods exposures.

Radon is considered to be responsible for causingpotentially 20,000 deaths in the European Unioneach year and 240 in Switzerland according to theSFOPH, being the second cause of lung cancer aftersmoking and more deathly than aids. The averageindoor radon concentration in Swiss living rooms isabout 75 Bq/m-3, but is oddly distributed throughSwitzerland (Figure 1).

Mapping the concentration of radon for the na-tional territory will be a valuable tool for decisionmakers related to public health and for people in-volved in house remediation. In the last year a largenumber of researchers have contributed to describeand measure the presence of gas at inhabited areas.The approaches came from the physical descriptionof radon behaviour starting from lithology as thegeneration source to the interaction with buildingvulnerabilities and the final risk due to human be-haviour.

Regarded as a pure natural hazard many investi-gations have focused on the description of indicatorvariables such as lithology, soil permeability, soiltemperature, air temperature, parental material – soilrelations, soil migration, sediment permeability,building material exhalation, ventilation and others(Medici, 1994; Strand 2005). The spatial modelling ofall the factors together remains unfeasible and oftenthe individual analysis results in loss of information,bias and not useful for predictions at local level(Miles, 1998). Indoor radon concentration of single

houses are not predictable from geological maps,because construction type and structural fabric ofhouses are essentially governing the extent to whichsubsoil radon potential affects the indoor concentra-tion (Kemski, 2001).

In the present research different approaches andhypotheses were applied to understand and to char-acterize spatial patterns. The multivariate nature ofthe radon indoor concentrations triggers many diffi-culties for monitoring, modelling and data visualisa-tion. A first hypothesis was proposed in the sensethat radon measurements could be considered effi-cient enough to explain spatial distribution.

The analysis is complicated by influences of fac-tors such as floor level and inhabited condition, roomtypes, etc. By removing geographic duplicates andselecting the maximum values a pessimistic scenariocan be considered. Finally spatial declustering andregularisation by means of moving windows statisticswas performed.

The spatial interpolation followed a sequence ofcross-validation to determine optimal lag and angleintervals during experimental variography. Variogra-phy modelling and sequential Gaussian conditionalsimulations were carried out to generate multipleequally probable realisations of spatial patterns. Postprocessing of derived digital spatial models givesrise to indoor radon and risk mapping.

Limitations were observed for homogeneous cov-erage of areas and for better levels of precision, as itwill be desirable to provide reliable information forunsampled areas and scales with better precision(1:25000 maps). Completion of gaps was tried tosolve by modelling correlation with existing informa-tion about geotechnical units. The big size of theunits and the weak correlation with indoor radon willnot contribute to improvements for clustered areasbut rather they introduce additional noise. It could beexpected that a fairly contribution at a regional scale

Efficient spatial predictors for indoor radon mapping

*Tapia, R., *Kanevski, M., *Maignan, M., **Piller, G. & **Gruson, M.

* IGAR, Institute of Geomatics and Analysis of Risk, University of Lausanne

** SFOPH, Swiss Federal Office of Public Health

(figure 2) could be established if soil permeabilityand/or outdoor measurements using gamma-rayspectrometers were available.

Also the Incorporation/assimilation of any re-lated/correlated information from geological orlithological origin can be carried out using machinelearning algorithms, e.g. artificial neural networksand support vector machines (Kanevski and Maig-nan, 2004).

Figure 1. Radon risk zones in Switzerland

Figure 2. Geotechnical map showing regional trends for radondistribution

REFERENCES

Medici, F. and Rybach L. (1994): Measurements of indoor ra-don concentrations and assessment of radiation. Journal ofApplied Geophysics, Volume 31, Issues 1-4, Pages, Feb-ruary 1994, 153-163.

Miles, Jon (2005): Development of maps of radon-prone areasusing radon measurements in houses. Journal of Hazard-ous Materials, Volume 61, Issues 1-3, August 1998, Pages53-58

Kemski J. et al (2001): Mapping the geogenic radon potential inGermany. The Science of the Total Environment 272, 217-230.

Strand, Terje et al (2005): High radon areas in Norway. Inter-national Congress Series, Volume 1276, February 2005,Pages 212-214

Wichmann, H.E. et al (2005). Lung cancer risk due to radon indwelling, evaluation of the epidemiological knowledge. In-ternational Congress Series, Volume 1276, February 2005,Pages 54-57.

Kanevski, M. and Maignan, M. (2004): Analysis and Modellingof Spatial Environmental Data. EPFL Press. 288 pp. (Ed.),Remote sensing for environment monitoring, GIS applica-tions and Geology. Proc. SPIE, Vol. 4545: 1-12